PulseExploreJournal ClubDebatesTrendingResearchersJournals
Instagram
HomeExploreJournal ClubTrending
Synapse
⌘+K
Synapse
May 6, 2026Journal of Nonlinear Complex and Data Science0 citations

Nonlinear thermal behavior prediction of biomimetic battery cooling systems using data-driven deep learning models

View Full Paper
AHA. Tawfeeq HussainKTK. ThavasilingamDD.Sakthimurugan

Key Points

  • This research aims to develop a biomimetic Battery Thermal Management System (BTMS) for electric vehicles, inspired by the Morpho didius butterfly.
  • Implemented a bioinspired BTMS design based on butterfly microstructure.
  • Conducted CFD simulations using different materials: graphite, copper, and aluminum.
  • Applied various machine learning models including deep learning approaches like CNN and feedforward neural networks.
  • Achieved prediction accuracy of R2 = 0.994 using CNN for temperature gradients and convective heat flux.
  • Significantly reduced CFD simulation iterations from 50 to less than 10 through surrogate modeling.

Abstract

Abstract A new bioinspired Battery Thermal Management System (BTMS) for electric cars is presented in this work. It is based on the microstructure of the wings of the Morpho didius butterfly. The thermal performance of a conical frustum fin design was assessed using transient thermal and CFD simulations in three different materials: graphite, copper, and aluminium. The biomimetic casing is combined with machine learning (ML) and deep learning (DL) models, such as XGBoost, Feedforward Neural Networks (FNN), Convolutional Neural Networks (CNN), and Linear Regression, in contrast to previous methods, to replace iterative simulation cycles. With a prediction accuracy of R 2 = 0.994 and RMSE = 0.098 °C, the CNN model was able to predict temperature gradients and convective heat flux under a variety of design conditions. The AI-driven system achieved design optimization speed improvements by decreasing repeated CFD computations through the replacement of about 50 iterative CFD design evaluations with less than 10 validation simulations after training the surrogate model. 3D plots, hybrid heatmaps, and residual maps are examples of advanced visualisation approaches. This work creates a scalable basis for applications in next-generation thermal systems by utilising CNN-enabled surrogate modelling for the first time on a Morpho-inspired EV case.

Ask AI
Helpful
Bookmark
Share
View Full Paper

Cite This Study

Hussain et al. (2026) studied this question.

synapsesocial.com/papers/69fa8e8904f884e66b530e12https://doi.org/10.1515/jncds-2025-0066
Ask AI
Helpful
Bookmark
Share
View Full Paper